Skip to main content



Executive Summary:
Water holding capacity is an important quality indicator in meat processing since affects both the yield and the quality of the end product. Poor water-holding capacity costs the meat industry millions of euros annually. The overall profit losses of the meat processing industry due to quality problems in pork meat processing is estimated to be over €5.1 billion by year.
QMEAT is a new sensor technology for the automatic classification of fresh pork meat according to its dielectric properties and colour. These parameters are related to quality attributes of the meat affecting meat processes like cooking and salting. The system has been designed to sort raw meat in different quality groups in meat processing lines. QMEAT represents a step forward in relation to the state of the art meat classification techniques as the proposed application is based in the use of fast, automatic and contactless technologies. The system is unique in the market as current systems for the determination of quality attributes are based on hand held invasive probes (pH and colorimeters) which are not well suited for in-line classification.
Q-MEAT advances over the state of the art:
• Automatic, objective and reliable sorting of individual pieces of meat in quality groups by the determination of dielectric properties and colour.
• System based in non-invasive technologies to avoid damage to meat and cross-contamination.
• Easy assembly on industrial pork cutting lines and capable to in-line classification of meat pieces at high speed (up to 600 pieces/hour). System easy movable and adaptable to automatic sorting lines already in the market.
• Affordable commercial cost with reduced maintenance.
• QMEAT can be used to identify producers providing higher quality raw meat.
• QMEAT sorting technology reduces the variation of salt content in dry-cured products.
• QMEAT sorting technology reduces cooking losses in the production of cooked ham.
• Customer satisfaction by improving quality and consistency of the end products.
• To monitor and reduce the real incidence of poor quality meats at slaughterhouse.

Project Context and Objectives:
In the EU, the most consumed processed meat products come from pig meat. The industry is dominated by many family owned companies running relative small – medium productions and involving traditional processes. These companies produce a wide variety of traditional quality pig meat products like cured and cooked ham, dry and fermented sausages, etc. Several SMEs export products outside the EU. However, these EU SMEs are facing increased global competition from large international companies which are more productive and efficient and have superior marketing organisations. Moreover, due to the global crisis, there is a continuous reduction in consumer demand for traditional premium quality products that is the core business for most European sector SMEs.
Only in the EU the sector concentrates a total of 45,000 SME companies working in the manufacture of processed meat products and employing about 1 million workers. Most of the SME meat processors in EU, still base product quality controls on human assessment or in random periodic analysis. This way of operation is not efficient and lead to significant variability in the final quality of the product and unsatisfactory productivity. A technological advance is necessary to overcome the lack of efficient technology to classify and sort pig meat by its quality prior processing.
SMEs will be more competitive if high quality raw meat is sorted for those processes that generates more added-value while lower quality raw meat can be treated to get an acceptable and consistent final quality. The sorting of raw material lets the producer: 1) to increase the industrial yield, 2) to obtain a more uniform product quality and 3) to identify those providers (slaughterhouses and wholesalers) that supplies higher quality raw meat.
The overall profit losses of the meat processing industry due to quality problems in pork meat processing is estimated to be over €5.1 billion by year.
The proposed solution:
Water Holding Capacity and colour are both key quality indicators in meat processing. Both parameters affects the appearance of the fresh meat and thus the consumers` perception at the time of purchase. WHC affects the technological value of the fresh meat and the juiciness of the processed product .
The processing of raw meat with inadequate WHC negatively affects industrial yield (e.g. weight losses and rejected product), inferior taste and juiciness, discoloration and spoilage of product. Colour often reveals insufficient (pale red) or excessive WHC (dark red), but this quality parameter is more critical for the marketing of fresh meat as the key deciding factor for consumers when selecting pork cuts at the point of purchase.

Present technologies to objectively determine WHC and colour are based in hand held invasive or contact probes (e.g. pH meters, conductivity meters, colorimeters) which are time consuming, contamination-prone, not suited for high speed on-line processes and not representative of the whole batch (a random sampling is typically performed).
QMEAT technology is a unique tool that will let meat processors to improve industrial yield, minimize product quality variability and to improve the value of lower quality raw meat. Additionally, the system is also very useful for slaughterhouses; especially for high speed pork cutting lines that split the whole carcass in the primal cuts as they will use data to select the best market destination (fresh market, further processing) and to quantify the real incidence of meats with inadequate WHC and colour. This information can be used by slaughtehouses to apply/suggest corrective actions at farm level and to improve pre- and post-slaughter processes at slaughterhouse level to minimize the production of poor quality meats. Several slaughterhouses have already shown interest in testing the technology.
The current lack of a commercial solution for this need is due mostly to the small and traditional nature of meat processors, an attitude which is now changing due to increased market pressure towards innovation and improved methods.
There are currently no sorting systems that provide the advantages of QMEAT technology: a non-contact and automatic meat quality classification system to objectively determine WHC and colour as principal technological quality parameters through measurements of permittivity and conductivity at multiple frequencies and colour. This will represent a great benefit to the quality and best practices of traditional European meat processing makers and the competitiveness of the overall sector.
Expected benefits for the SMES: In the EU, the incidence of inferior quality pigmeat varies by country and slaughterhouse, but is estimated on average to be around 10-15% for highly exudative meat (low WHC) with pale colour (Pale-Soft-Exudative meat), 40-45% for moderate exudative meats (medium-low WHC) but acceptable colour (Red-Soft-Exudative meat) and 5% for meats with excessive water retention (high WHC) and undesirable dark colour (Dark-Firm-Dry meat) , . The rest of meat is RFN (Red-Firm-Normal), which is the most appreciated class because, has optimum WHC and colour characteristics.
Technological problems in the processing of pig meat products:
Cooked hams: The production of cooked ham, made from pork meat with insufficient WHC, and which is carried out without the injection of artificial ingredients (e.g. phosphates, carragenates, artificial colorants, etc.) leads to textural problems provoking holes, muscle that breaks or crumbles easily which is very detrimental to the presentation of the product. In high speed slicing lines, there are large quantities of slices that must be rejected or are destroyed due to textural problems. The use of phosphates in cooked ham is typical and produces a significant increase in water holding capacity and improves inter-muscular binding which is very important for the slicing of the product. However, there is a continuous demand for products processed with natural ingredients.
Cured products: In cured products like ham, loin and bacon, insufficient WHC produces an important reduction in weight loss, between 2-10% more than meat with optimum WHC, the absorption of salt is irregular (some parts are salty) and the product has excessive desiccation on the surface, obtaining a product of inferior quality. Moreover, excessive WHC or DFD meats will spoil more rapidly than normal meat and the product must be rejected at the end of the curing process.
Dry sausages: In the EU, producers of dry sausages reject up to 10% of the production due having low or excessive WHC in the mixture, while 30 % of the final sausages are poorly dried having a lower market value.
QMEAT addresses the following fundamental scientific objectives
• To investigate the use of innovative characterisation technologies for assessing the technological quality of meat, and its applicability to industrial operation
• To further the knowledge of the electrical and optical properties of raw pig meat
• To establish correlation models between the electrical and optical properties of pig meat and intrinsic meat quality parameters (Water Holding Capacity and colour)
Industrial and economic objectives
• To achieve a manufacturing target cost for a complete QMEAT system below €30k.
• To optimise production by increasing industrial yield and producing more high quality products.
• To assist slaughterhouses to know their real incidence of poor quality meats and to determine the cause/s of the problem. QMEAT technology is expected to contribute in reducing the incidence of PSE meats and DFD meats at slaughterhouse.
The inclusion of QMEAT system will allow objective, non-contact and automatic classification of different pork meat pieces. (The yearly cost of manual analysis is estimated to be in the range of €20-25/tone, including the cost of probes and operator time). The cost of QMEAT analysis will be 1.1€/t assuming a lifetime of 10 years.
Societal Objectives
• To achieve customer and consumer satisfaction and confidence by the production of consistent quality products.
• Increase of direct employment in QMEAT manufacture, installation and maintenance services. Additionally new job opportunities will be created, especially for younger people in rural areas for tasks where previously years of experience were required.
• Improving the working conditions of the employees by the use of an automatic classification system.
Contribution to Community Society objectives:
QMEAT will contribute to the continued growth of the EU pig meat sector as well as to increase the trust of EU consumers by providing consortium SMEs, and eventual end-users, with a system that accurately determine the technological quality of the meat before processing, resulting in a better final product with regard to its sensory features.
The work programme, as well as the potential benefits of the project, will provide collaborative opportunities for end users SMEs of the consortium and standardize an important aspect of production. Consequently, QMEAT will be more effective and increase the knowledge base of consortium SME producers, provide them tools to expand their businesses, and support employment growth in the industry. Furthermore, QMEAT will assist in competing against tougher competition from non-EU countries and “New Wold” countries like USA, Brazil and Canada.

2.Hanne C. Bertram, Jette S. Petersen, Henrik J. Andersen, Relationship between RN- genotype and drip loss in meat from Danish pigs, Meat Science 56 (2000) 49±55
3.Fournaise and Davies, (2003), Evaluation dans deux secteurs majeurs du commerce des viandes’’ et ‘‘Besoins en terme de spécifications pour la viande et conséquences de la qualité ou de la non qualité.
4.Gispert,M. Faucitano,L. Oliver,M.A. Diestre, A. (2000), A survey of pre-slaughter conditions, halothane gene frequency, and carcass and meat quality in five Spanish pig commercial abattoirs, Meat Science 55, 97-106.

Project Results:
Water holding capacity is an important quality indicator in meat processing since affects both the yield and the quality of the end product. Poor water-holding capacity costs the meat industry millions of euros annually.
Despite the relation between water retention and its dielectric properties and colour has been investigated and established, conventional measuring methods based on contact probes (conductivity, pH, colorimeters) are difficult to integrate for in-line product inspection and process monitoring. Major drawbacks of systems based on contact probes include: (1) poor accuracy, due to the sensitivity of the measurement to the surface contact resistance of the probe, and the positioning of the electrodes in the sample, (2) erosion and degradation of needle electrodes, (3) risk of bacterial cross-contamination, (4) low analysis speed, and (5) limited flexibility to inspect different types of products.
As an alternative to contact probe based methods, QMEAT technology offers the possibility to scan the whole meat part automatically and without contact by means of electromagnetic radiation and video image analysis. Therefore, the main scientific objective of QMEAT was to investigate the use of innovative characterisation technologies for assessing the technological quality of meat, and its applicability to industrial operation. The proposed technologies in QMEAT are Video Image Analysis to determine meat colour and Dielectric Spectroscopy to determine dielectric characteristics of the meat.
In the first period, from month 1 to month 9, the main objective was to validate the proof of principle of DS and VIA technology for predicting Water Holding Capacity and colour. To this end, two laboratory prototypes (DS and VIA) were designed and developed. To validate the proof of principle, it was decided to carry out experiments with pork loins. There are two advantages in using loins, 1) the tunnel aperture tunnel of the DS system is smaller and this simplifies the design and 2) Unlike other meat parts, the loin is a single muscle and this simplifies the study of WHC and colour.
WP1 - Dielectric Spectroscopy (DS) – Magnetic Induction Spectroscopy (MIS)
Dielectric Spectroscopy in biological tissues can be made by means of Electric Impedance Spectroscopy (EIS) or by means of Magnetic Induction Spectroscopy (MIS). Unlike EIS, MIS based systems do not require a physical contact with the object. Typically, in a MIS system, a variable radio frequency magnetic field is generated inside a tunnel. When the meat sample enters inside the tunnel, a primary magnetic field is induced in the meat sample. The induced magnetic field generates circular currents in the lean tissue (eddy currents) which generates a secondary magnetic field which is modified by the conductivity and the permittivity of the lean tissue. This secondary magnetic field is picked up by a receiving coil. Fat and bones can be considered transparent to the magnetic field.
A literature survey was made to determine the range of frequencies which better describes the relation between water retention and the dielectric and magnetic properties of biological tissues. This information was necessary to form a foundation to the development of a magnetic induction spectroscopy instrument to assess the quality of meat. Some studies carried out at experimental level with impedance probes (EIS), showed that the optimum frequency range is from 10 kHz – 1,000 kHz. According to the studies, the low frequency range appears to be more sensitive to Water Holding Capacity variations.
Once the frequency range was stablished, a coil configuration was proposed based on an axial gradiometer configuration. This configuration was chosen to minimise the dynamic range requirements for the sensors front-end electronics. An electromagnetic modelling of the sensor was developed to determine a suitable geometric configuration and its sensitivity.
Different electromagnetic solver packages were utilised to ensure the results were consistent in the whole range of frequencies (80 kHz - 1,200 kHz). The results indicated that the dielectric properties of the meat can be extracted for the proposed range. A further conclusion obtained from the simulations indicated that without the provision of electric field screening, the direct ‘capacitive’ coupling associated with both the excitation and sensor coils contaminates the magnetic coupling measurements.
Taking into account results obtained from the EM simulations, a laboratory test-rig was developed, implemented and tested at laboratory scale at IRTA (Girona, Spain) with fresh pork loins. The system includes:
1) A coil system or sensor head based in a gradiometer configuration (2 emitters and 1 receiver). The coil system has been properly shielded to minimise susceptibility to external magnetic fields and the effects of parasitic capacitive coupling.
2) A Power box including different elements to operate the MIS system, including electronics, control system to operate the conveyor belt and security systems (emergency stop and circuit breaker). The electronics was designed to work in a wide range of frequencies [80 kHz to 1 MHz].
3) A master box that includes a data acquisition system to acquire the In-phase and Quadrature signals coming from the power box and the signals from the IR photo-detectors. 4) Conveyor belt, motor and safety systems and 5) Host PC and Power box.
Before testing with real samples of meat, the system set-up has been characterised and calibrated by using phantoms with known dielectrical properties. The test rig can be programmed to work at 8 different frequencies in automatic multi-pass mode which means the sample is passed through the machine 8 times.
UNIMAN has been in charge on laboratory test-rig development. JCB has provided technical advice and follow-up design activities for the MIS sensor unit. UNIMAN has trained JCB in the use and assembly of the system. JCB was present at IRTA to install the MIS system and to train the researchers in the use of the system.
WP2 – Video Image Analysis (VIA)
ATEKNEA in collaboration with JCB was in charge to select main components of the VIA system including CCD colour camera, light source and computer for image processing. Different light sources, LED bars and fluorescents, were investigated. The components were assembled inside a measurement chamber which was designed for test at laboratory. After the assembly, the VIA system was characterized and calibrated using a colour reference chart.
A software package was specifically developed to operate the system. The software includes different parts: 1) System setup, 2) Acquisition of images, 3) Algorithm for calibration, 4) Algorithm for image segmentation and 5) Algorithm for colour space conversion (RGB to CIELAB). Algorithms were first implemented in Matlab® and later on in LabVIEW to automate the experiments.
The VIA system incorporates a module to determine the size and area of the meat parts. This parameter can be used to interpret data from the MIS system. Different technical solutions were discussed. Finally, it was decided to use a commercial solution based on a 3D sensor which has a cost of €700 approx. The device can afford the speed required for the application.
After the selection of main components forming the VIA system, a laboratory test-rig has been designed. The test rig includes all the necessary equipment to determine colour and size for any type of pork primal cut. ATEKNEA has been in charge on laboratory test-rig development. JCB has provided technical advice and follow-up design activities for the VIA system. Also JCB has provided support to ATEKNEA during the evaluation of the first VIA system proposing design improvements. A first test of the system has been made at IRTA together with the MIS system developed in WP1.
The first version of the VIA system includes 2 doors to put and remove the meat piece manually. However, the industrial version of the prototype includes a conveyor belt to allow the automatic pass of the meat through the chamber.
WP3 - Electrical and optical characterisation of meat at laboratory scale (first period)
The objective in WP3 is to predict the meat quality parameters based on the dielectrical (MIS) and visual (VIA) characterisation.
- Experiments carried out with loins
A first experiment was carried out with 26 fresh pork loins with a water holding capacity ranging from 1.75% to 16%, pH values ranging from 5.37 to 5.89 and Lightness (L*) ranging from 47 to 65 (Minolta CR-410).
• VIA system vs. colorimeter: Lightness (L*) was measured on 56 meat samples by the VIA system and by the colorimeter (CR-410). The correlation index achieved was R2 = 0.97 RMSE=0.74.
• VIA vs drip loss: Correlation models obtained from the first test with 26 pork loins using L* and a* as a variables, provides good prediction accuracy for the estimation of % drip loss: R2= 0.863 RMSE= 1.5.
• MIS vs drip loss: Correlation models using MIS signals at 40 kHz and 100 kHz, obtained from the first test with 26 pork loins, provides reasonably good prediction accuracy for the estimation of drip loss, Adj. R2 = 0.75 and RMSEC=2.0%.
• MIS+VIA vs. drip loss: Twenty six pork loins were scanned with the DS and VIA systems. Drip loss were determined by traditional methods. The correlation index obtained was R2 =0.936 with RMSEC=1.0 RMSEV=2.4.
- Experiments carried out with hams using VIA system to determine Lightness (L)
An experiment was carried out with 80 fresh hams to determine the accuracy of the VIA system in predicting lightness (L*). The reference method used was the Minolta CR-410. Three parts of the ham were selected to analyse the colour: Semimembranosus (under Gracilis), Semimembranosus, and Abductor. The best correlation was found in the muscle Abductor, R2=0.85 RMSE=1.57. Moderate correlations were found for the other two muscles: Semimembranosus R2=0.69 RMSE=2.38 and Semimembranous under Gracilis R2=0.53 RMSE=2.61. The lower correlation found in the Semimembranosus can be explained due to the presence of fat deposits. One of the main advantage of the VIA system respect traditional colorimeters is that fat deposits and connective tissue can be removed from the image.
- Experiments carried out with hams using VIA system to determine Lightness (L)
An experiment was carried out with a selection of 18 fresh hams that were cooked by partner RDVIC. The selection of the ham was made according to different pH ranges (high and low). The traceability of 6 hams was lost during the cooking process and only 12 hams were analysed and compared with the L*a*b* data provided by the VIA system. The average cooking loss was 4.6% with a variation of 1%. Using L* and b* and variables, the correlation index found in this experiment was: R2=0.79 RMSE=0.48. Despite the low number of samples used, the results from the experiments indicated the potential of the VIA system to predict % cooking loss.
- Experiments carried out with hams using VIA and MIS system
Raw hams were characterised to predict technological meat quality and cooked hams were elaborated. All response variables evaluated in raw and cooked hams: raw ham drip loss, ham cooking loss, slice drip loss and slice denaturation score were significantly affected by the pH group effect and showed negative non-linear relationships with pH measurements. The highest linear correlations of ham drip loss (-0.46 to -0.54) ham cooking loss (-0.43 to -0.50) slice drip loss (-0.46 to -0.54) and slice denaturation score (-0.65 to -0.70) were observed with pH measurements at different positions.
Magnetic Induction Spectroscopy (MIS) parameters, quadrature at 40 and 80 kHz and quadrature ratios 320/40, 640/40 and 320/80 showed the highest significant differences according to pH group. Regarding Video Image Analysis (VIA) system, colour parameters of ham position M5 showed the highest significant differences according to pH group.
The best technological meat quality predictive models obtained by Stepwise Regression analysis using non-invasive measurements (MIS, VIA, weight) were for raw ham drip loss with 58.9% of the response variable explained and for cooked ham slice denaturation score with 48.7% variation explained. When invasive measurements were included in the models (pH measurements), the percentage of variation explained increased slightly in the case of raw ham drip loss (60.1% variation explained) and this increase was much higher for slice denaturation score (62.2% variation explained). Regarding cooking loss and slice drip loss, the best predictive models obtained explained 44.7% and 42.1% of variation, respectively, by combining invasive (pH) and non-invasive measurements.
PLS regression improved slightly the predictive models for ham cooking loss (49.5% explained variation; Prediction RMSE=1.000%), slice drip loss (46.3% explained variation; Prediction RMSE=0.787%) and slice denaturation score (64.0% explained variation; Prediction RMSE=0.985) using both invasive and non-invasive measurements.

WP4 – Industrial design and development of the QMEAT system
One of the objectives of this WP was to scale up the MIS system to validate the technology for other meat cuts such us hams, fiocco, neck, etc. To this end, a MIS system has been developed to carry out extensive experiments at laboratory in WP3 and incorporates a number of significant improvements respect the first MIS unit developed for loins. Among the most remarkable improvements: 1) Tunnel aperture (60 x 60 cm) to allow the pass of large meat cuts, 2) Multi-tone signal (up to 8 frequencies) to scan the meat in one single pass, 3) Scanning time (600 samples/hour).
After validation at laboratory scale (WP3), a pre-industrial version of the QMEAT system has been made. The two modules (MIS and VIA) forming QMEAT system have been conceived to work independently. This solution offers a number of advantages including lower manufacturing cost, flexibility, particular implementation of VIA/MIS technology for other applications, etc. Other factors that has been taken into account for the industrial design includes; easy installation, maintenance, mechanical protection, hygienic design and encapsulation.
The main results of WP4 is the development of an industrial version of the MIS and VIA system and the calibration with different cuts of meat (ham, loin, fiocco, neck, etc).
Result 1: Industrial version of the MIS system to determine dielectric properties (conductivity and permittivity) in meat cuts in a wide range of frequencies [80 KHz – 1,200 kHz]. The system is suitable for in-line operation in a meat processing plant and works at 600 pieces by hour. The system has a protection index IP65, incorporates a dynamic weighting system with an accuracy of +/-10g, a bar code reader a touch panel and Ethernet connection. The manufacturing and assembly costs of the system including the conveyor belt and weighting system is 17.5k€.
Result 2: A Video image analysis system to determine colour (L*a*b*) in meat parts. The VIA developed can be a real alternative to the traditional hand held colour probes. The system is suitable for in-line operation in a meat processing plant working at 600 pieces by hour. The VIA system has been designed to provide a protection index IP65, suitable for an industrial environment. The manufacturing costs of the system including the conveyor belt is 10.7k€.
Results 3: Pre-calibration of the MIS and VIA system with a significant number of meat parts. The multivariate analysis of QMEAT signals allows to sort meat parts (e.g. loins, hams, coppas, fioccos, etc.) in terms of their technological quality (colour, attitude to salt uptake and cooking loss). As a rule, fresh cuts with higher amplitudes of MIS signals and high L values in VIA had lower WHC, higher drip-loss, higher cooking-loss and higher salt intake. In the case of fresh cuts with high fat amount like coppa or fiocco, samples discrimination resulting from QMEAT application is affected by fat content.
WP5 – Integration at Industrial level
Task 5.1 of WP5 was dedicated to assemble the QMEAT prototype for working in industrial conditions and to test it with fresh cuts prior to processing. According to main field of production of SME San Michele, QMEAT was tested with fresh coppas and hams coming from Italian heavy pig. Coppas and hams were hot sectioned from carcasses, refrigerated for 24h and delivered to the plant by refrigerated carriage. Both coppas and hams were supplied by two slaughter houses. After unloading, fresh cuts were moved to conditioned rooms for tempering at low temperature (1 – 3°C) before processing.
1) QMEAT scan of fresh cuts
A total of 100 coppas to be scanned with QMEAT were selected in the weight range 2.5-3.0 kg. Visual colour (with main reference to dark muscles of the cut) and pH were the selection criteria: pH was measured in two coppa positions (dark and pale muscles taken as reference), using the pH-meter 330 WTW by electrode insertion. Due to the shape of fresh coppa, fat content was not detectable by outer visual evaluation. Before sample scan, the system performance was checked by saline solutions (1-3 S/m), to be able to match signals achieved in different working sessions. Fresh coppas were scanned at fixed temperature (2-3°C). All coppas underwent QMEAT scan to measure color indices by Video Image Analysis (VIA) and dielectric properties by Magnetic Induction Spectroscopy (MIS). At industrial level, VIA signals were acquired on dark muscle. A total of 9 signals were acquired as color indices L*, a* and b* (VIA), and amplitudes (A) by MIS after the application of 6 radio frequencies ranging from 79 to 1,120 kHz. To overcome variability related to coppa weight the amplitudes were corrected for the weight of the scanned samples.
Raw hams were chosen in the weight range 12.4 - 13.7 kg; 50 hams variable in pH (measured on semimembranosus muscle, SM) were selected. Visual colour (with main reference to muscles semimembranosus and biceps femoris), pH and subcutaneous fat thickness (measured with a ruler calibrated in mm), were the selection criteria. pH was measured using the pH-meter 330 WTW by electrode insertion in SM. The established subcutaneous fat thickness range was 25-33 mm, i.e. in agreement with fat thickness recommended by Parma ham tutelary regulations. QMEAT scan was performed in the same way than in the case of coppas, and to overcome variability related to ham weight, the amplitudes were corrected for the weight of the scanned samples.
2) Dry-salting procedures
After QMEAT scan both coppas and hams underwent dry-salting. Coppas were salted according to the standard procedures of San Michele (SANMI): two-steps salting stages with a mixture of salt, spices, nitrites and nitrates. Percent added salt (2.7% including both salting stages) was calculated according to trimmed coppa weight. The salting treatment lasted a total of 7 days storage in a cold room, at temperature 2-4 °C. Hams were salted according to the standard procedures of San Michele (Parma ham producer): two salting stages, both with dry and wet salt. Percent added salt (6.4% including both salting stages), was calculated according to trimmed ham weight. The salting treatment lasted a total of 19 days storage in a cold room, at temperature 2-4 °C and relative humidity > 85%.
At the end of the salting stage, coppas were weighed to calculate the % salting loss, expressed as percentage of fresh coppa weight; next, a representative number of samples were dissected to calculate absorbed salt, moisture and total fat content, while a part was kept to accomplish the full processing.
At the end of the salting stage, hams were weighed to calculate the salting loss%, expressed as percentage of fresh ham weight, and a representative number of hams were dissected to calculate total fat and lean content, while a part was kept to accomplish the full processing. Subcutaneous fat with rind and muscles with inter and intramuscular fat (muscular part) were separated. The parts were weighed and the muscular parts were analyzed for fat content. Total fat and lean weights were expressed as percentages of salted ham weight. The lean part was analysed for salt (NaCl) and moisture contents. Both in coppas and hams, moisture, fat and salt were expressed as g/100g wet muscle; salt was also calculated as % NaCl dry matter (g/100g dry matter) to overcome differences in moisture content, and as % NaCl muscle fluids (% NaCl / (% NaCl +% Moisture)), to report the concentration of muscular saline solutions.
Data Analysis. A multivariate PCA model was calculated, based on QMEAT signals taken at SSICA from the 10 tested coppas and at SANMI from the first batch of 14 coppas; these scanned samples were assayed for pH and fat content. The same experimental design was applied to hams (9 from SSICA and 9 from SANMI). After matching signals collected in different places and properly aligned, the objective was to check if:
1) a single model could account for most variance associated to selected assayed samples scanned at lab-scale (SSICA) and at industrial scale (SANMI);
2) Unknown coppas and hams scanned at SANMI with QMEAT can be classified on this model.
The Unscrambler statistical package ver. 9.8 was run, to accomplish data analysis including the unknown sample classification; the multivariate model was validated by the option “Full Cross Validation”. The amplitude corresponding to the lowest frequency (f =79 kHz) was excluded from data analysis because of the low signal-to-noise ratio. Concerning VIA variables, all data recovered from SSICA and San Michele were normalized before statistical analysis, to merge data acquired before and after the replacement of VIA camera, caused by a technical problem occurred after the transfer of VIA module from SSICA to SANMI.
Results. 1).Two principal components (PC1 and PC2) were computed, accounting for 63 and 19 total variance respectively (% explained variance = 82) for coppas and 72% and 17% total variance respectively (% explained variance = 89) for hams. The results confirmed, in agreement with the models obtained at SSICA at lab-scale level, that samples scored on PC1-PC2 plane, can be soub-grouped according to amplitudes and VIA color indices, obtaining groups differing in fat and salt content (coppas and hams), and pH (hams). After QMEAT scan, the information is that samples with higher amplitudes and higher LVIA (group 1), are associated to lower pH (PSE/RSE type), lower fat content, higher proneness to salt uptake than samples with lower amplitudes, lower LVIA higher aVIA (group 3). Samples with intermediate properties and scores were put in group 2. Out of 24 coppas from SSICA and SANMI, 10 achieved scores meeting group 1, 5 were in group 2 and 9 were in group 3. In the case of hams, 7 were in group 1, 6 in group 2 and 5 in group 3.
2) The remaining 86 “unknown coppas” and 41 “unknown hams” were scanned by QMEAT in different sessions; signals achieved were used as data set in Unscrambler, to be classified on the previous models (one for coppas and one for hams). When projected onto the PC1-PC2 plane, unknown samples were scattered across the plane, with scores included into the bounds given by samples used for the classification models. This occurrence shows that, even if scanned in different sessions and coming from different suppliers, ungrouped unknown samples are evenly distributed and fit a common model. According to these findings, QMEAT prototype, can be used to classify raw meat after supplying, and the classification responses are time- and place independent; signal alignment by means of saline solutions and calibration of MIS and VIA devices are needed to make comparable samples scanned in different sessions.
Task 6.1 of WP6 was dedicated to validate in industrial conditions the performance of QMEAT system at SANMI in classifying fresh coppas and hams prior to processing. In task 5.1 a PCA model was elaborated, made of both samples tested at SSICA and first samples tested at SANMI, identifying three groups for each type of fresh cut (coppa and ham), based only on QMEAT signals, and confirmed by analytical measures (pH, fat content and absorbed salt). Next, all unknown fresh samples scanned at partner SANMI were classified onto PCA plane, to check if new ungrouped samples fit the model. According to the obtained positive response, in task 6.1 scanned samples were:
1) Classified into groups differentiated for meat and technological quality;
2) Validated for group membership
Reference groups for classification are characterized as follows:
- 1: group of samples with QMEAT response based on high amplitudes and high L values, low pH24h, low aVIA and/or low fat (including cuts showing PSE/RSE-type properties);
- 2: group of samples with intermediate QMEAT response and analytical values (including cuts showing normal/RFN-type properties);
- 3: group of samples with QMEAT response based on low amplitudes and low L values, high pH24h, high aVIA and/or high fat (including cuts showing DFD-type properties).
After scan, all coppas and hams underwent dry-salting; at the end of salting phase, a representative number of samples for each classification group was dissected to be analyzed for fat, moisture and salt content (validation of assignment to groups). Remaining samples were left to finish the dry-curing processing.
Data Analysis. The whole data set (110 coppas and 66 hams) underwent Multivariate Discriminant Analysis (SPSS statistical package ver. 19.0): 24 coppas and 18 hams previously used for the PCA model were labeled as training sets, while the unknown ungrouped samples scanned at SANMI with QMEAT system were labeled as test sets.
Results. Discriminant functions were calculated for group 1, 2 and 3 (training sets), based on QMEAT signals: next, all scanned coppas and hams were classified and assigned to groups 1-3, according to discriminant functions, including samples used as training sets, that were reclassified. In the case of coppas, assignment to groups was statistically significant for 86% of samples; remaining samples were classified according to the highest probability of group membership. All cross-validated cases were classified according to previously predicted groups. Among scanned samples, 49 coppas were assigned to group 1, 47 to group 2 and 14 to group 3. In the case of hams, assignment to groups was statistically significant for 85% of samples; remaining samples were classified according to the highest probability of group membership. With the exception of one ham (classified in group 3 and ascribed to group 2), cross-validated cases were classified according to previously predicted groups: within scanned samples, 14 hams were assigned to group 1, 33 to group 2, and 19 to group 3.
At the end of salting phase (lasting one week for coppas and 19 days for hams), 55 selected samples, analyzed for pH, fat and salt content, were used to validate classification and assignment to groups. Ham grouping was successful for quality and technological response: scanned fresh hams classified in group 1 (high amplitudes, high LVIA and low aVIA values), at the end of the salting phase have the highest salt (NaCl) uptake, the highest salting loss, the lowest pH and the lowest fat content; fresh hams classified in group 3 (low amplitudes, low LVIA and high aVIA values), at the end of the salting phase have the lowest salt (NaCl) uptake, the lowest salting loss, the highest pH and the highest fat content; fresh hams classified in group 2, with intermediate QMEAT values, give salted hams with intermediate values. From a numerical point of view, the differences are moderate though significant (P<0.05) and able to strongly increase variability in final outcome. For instance, in the case of salt uptake, scanned hams assigned to group 1 and analyzed for validation had % NaCldry matter = 11.1 vs % NaCldry matter = 9.9 of hams assigned to group 3. Hams assigned to group 1 were also leaner and more prone to weight loss, as also witnessed by the lower pH: it means that, final product is expected to yield a remarkable concentration of absorbed salt (that is higher after salting). On the contrary, hams assigned to group 3 are fatter and with low weight loss, delaying the concentration of absorbed salt (lower than in group 1). As long as fresh hams are mixed together and processed in the same way, the combination of these traits increases the variability of final batch: otherwise, if groups of fresh hams generated by QMEAT classification are processed in a specific way coming from their properties, final products can result more homogeneous.
Even in the case of coppas, QMEAT classification into groups accounted for differences in fat content and salt intake, matching the same comments presented for hams; for coppas, classification did not produce significant differences in pH between groups, even if the average values of pH in groups are in agreement with the general trend, i.e. group 1 has lower pH than group 3 (approaching pH = 6.0 in dark muscle). The better results achieved for ham than for coppas, could be ascribed to QMEAT arrangement, currently more appropriate for ham morphology and size than for coppas. In addition, red muscles seem to prevail over the white ones in coppas, whereas in ham, white muscles are predominant: may be that in white muscles water status and dielectric properties are more affected by pH variation than in red muscles. However, it is known that coppas with high pH in red muscles (pH ≥ 6.0) and high fat amount have serious problems in texture and taste: in this respect, the possibility to identify them by QMEAT inspection could be a tool for improving homogeneity and quality of final outcome.
Finally, mainly for hams but also for coppas, the selection of raw matter based on QMEAT signals is effective and time saving if compared to procedures based on the measure of traditional laboratory analyses allowing the classification into PSE/RSE, DFD and normal meat.
The task 6.2 aims to evaluate and discuss among all members of the consortium data obtained during testing and validation of QMEAT in order to determine if QMEAT module meets their expectations.
The QMEAT prototype is based on contactless Magnetic Induction Spectroscopy (MIS) and Video Image Analysis (VIA) technologies; the main goal is to establish correlation models between the electrical and optical properties of pig meat and intrinsic meat quality parameters (water holding capacity and color), influencing meat response to processing, i.e. technological quality.
To test this possibility, loin was used as a model system: it was trimmed to maintain a single muscle (Longissimus dorsi) with a regular shape, deprived of covering fat, leaving marbling fat (% range = 1-5 on wet muscle), as the lonely source of variability between samples. In this respect, pH, water holding capacity and color, regarded as the parameters defining meat quality in terms of PSE, RSE, RFN and DFD and having an impact on QMEAT signals, are unaffected by the interference of other issues like fat amount and distribution, shape, presence of different muscles with a different metabolic pattern, as occurring in complex cuts like ham or coppa. In data analysis carried out at IRTA, it was possible to build predictive regression models of water holding capacity (expressed as % drip loss) of loins, using QMEAT outputs and derived variables. At SSICA, loins were classified into groups obtained by multivariate scores and based on principal components calculated with QMEAT signals: loin groups generated by QMEAT signals differed in color, water holding capacity (WHC) and salt uptake. Both approaches established that, in the case of loin model, QMEAT measures were effective in giving information about loin technological properties both in terms of prediction and classification. The small MIS unit developed for loins was employed to accomplish loin scan and signal acquisition.
Other investigated cuts, representative of raw matter normally used by meat processors to be manufactured into meat products, i.e. hams for dry-curing and cooking, coppas and fioccos for dry-curing, were scanned by the big MIS units, one provided by UNIMAN and working at the facilities of Spanish partners, and one provided by JCB and ATEKNEA and working in Italy.
In the case of cooked ham response variables like raw ham drip loss, ham cooking loss, slice drip loss and slice denaturation score, resulted affected by pH, but not properly predictable by means of QMEAT signals. Otherwise, in the case of fresh hams undergoing dry-salting or cooking, QMEAT response was effective for classifying fresh cuts according their technological quality, distinguishing fresh samples into groups, that after processing differed in cooking loss (for cooked hams) and salt intake (for dry-salted products). The different proneness of fresh cuts to cooking loss and salt intake, was supported by combinations of pH, WHC, fat content and colorimetric indices, and estimated by a multivariate combinations of QMEAT signals. Statistical software as PCA and Multivariate Discriminant Analysis were effective to define the training sets and to classify scanned samples into groups. The influence of fat content on QMEAT response (MIS signals), was demonstrated; the classification based on QMEAT response could be also complicated by weight differences and by the presence of several muscles (white and red muscles). To overcome these problems, hams were selected in a narrow range of weight and visible covering fat: a similar preliminary grouping in weight and fat classes can be performed in plants, before QMEAT scan. Next, amplitudes of MIS signals should be corrected by the weight of each cut, to avoid difference in signals due to size. In this respect, the QMEAT system should be equipped with a scale. As to VIA, a second camera can be requested to provide the image of other regions of the cuts, in addition to the zenithal one.
In the industrial application of QMEAT prototype, fresh pork cuts classification can be very useful to meat processors, to reduce variability in salt content, fatness and texture (a property strongly affected by pH, salt and fat) in final outcome, selecting the raw materials more appropriate to plant processing way or adapting the processing way to raw matter according to QMEAT response.

Potential Impact:
The pigmeat sector in Europe
In 2013, the most consumed meat in the world was pigmeat with 104 million tons (mt) while chicken was 73mt and beef 52 mt. The EU27 is the second largest pig meat producer in the world with a 21.6% of market share (22.3 mt), only surpassed by China with 45.7% of the global production. In 2013, the EU25 was 111% self-sufficient and the imports were insignificant at only 16 mt. However, the EU-27 as a whole is losing its share of the world market since 2008. These losses in global market share have been picked up by other competitors like the USA, Canada and Brazil. The long-term competitiveness of the EU is not very promising , due to animal feed cereal constraints (e.g. biomass competition for arable land), and costs associated to animal welfare laws (Directive 2001/88/EC) and environmental regulations (Directive 2008/1/EC). The rising feed costs have put the sector under pressure during the last years.
In the EU25, pig meat is the most consumed product (20.1 mt during 2013), compared to beef/veal (8.1 mt). Within the key producers, Germany (22.6 %), Spain (15.4 %), France (10.1 %), Poland and Denmark (7.6 %) produce more than half (55.7 %) of the pig meat in the EU and concentrates more than 39,165 meat processing SMEs . In 2013, these countries produced about 18 million tonnes of processed meat, representing 83% of total European production.
The meat sector plays a very important role in the economy of Europe in terms of labour and output value, as well as in the development of rural areas. In 2012, the meat production accounted for a 9.1% share of the EU-27 overall agricultural industry output (more than €32.87b in profit). The employment, in full time equivalents, is estimated at 1 million workers . The industry is very fragmented and dominated by small SMEs and characterised by a complex network of farmers, farmer cooperatives, slaughterhouses, processing and rendering plants and retailers. According to CIAA , the meat processing sector generated a turnover of €160.6 billion in the EU-25 from which approximately €85 billion were by SME companies.
Beyond global competition, EU sector SMEs are faced with falling consumer demand for premium meat and meat delicacies, declining since 2009 due to the economic crisis. The more expensive the meat, the bigger the fall in demand, as consumers move to lower price options . In terms of product price, large pig meat manufacturers are more competitive than SMEs, as they have bigger and more efficient plants, more technological resources and superior marketing organisations. Large companies usually control pig growing farms, slaughterhouses, cutting lines, packinghouses and meat processing plants. This degree of integration, automation and control allow them to be very efficient, to reduce costs and to increase the quality of the produced raw meat and thus the quality of the final product.
In the competitive global market it is essential for SMEs to increase their competitiveness by optimising their production processes (e.g. to increase industrial yield) and produce consistent quality products (to raise the standards of quality). Due to the above mentioned factors, numerous sector SMEs are going out of business and their market share is directly picked up by large companies. Given the expected slow recovery of the EU economy, the survival of the major number of SMEs will be a key issue for the consolidation of the sector.
Economic impact and SME competitiveness
The QMEAT technology will result in significant economic benefits to EU meat processing SMEs, who will benefit from a technology that will let them to increase industrial yield and achieve higher added value-added end products with greater consumer appeal. On the other hand, EU abattoirs will also benefit from the QMEAT technology. Meat exudation and unacceptable colour are the main factors affecting the marketing of fresh raw meat and contributes to significant economic losses because the value of the cut is decreased. In the EU, near 152 million pigs were slaughtered in 2013, with a cost due to PSE for abattoirs estimated at €453M . The information provided by QMEAT will help abattoirs identify producers that provide poor-quality meat and improve processes to reduce and control the incidence of poor quality meats.
By implementing QMEAT system, meat pieces will be separated by WHC and colour. Those pieces with better water retention will be destined to higher added value processes (long curing processes, natural added ingredients, etc.). Especially in the production of cured ham and loins, a reduction by 20% for the “budget” quality is expected. Major part of the “budget” quality is produced from hams imported from other European countries that do not achieve standard quality. An exhaustive control quality of those meat pieces at arrival will allow selecting better providers of raw meat products.
The market
The main potential markets have been identified as meat processors and slaughterhouses. The product addresses these potential markets, as a branded product to be incorporated into existing pork processing/cutting lines.
Regarding the potential market, several sources, such as CLITRAVI, FENAVIAN and the AICE , establish that in 2012 the number of dry pig meat manufacturers SMEs (>20 employees) was about 20,000 and 3,300 abattoirs while Eurostat , estimates that that Spain has about 10% of the companies involved in production, processing, and preserving of meat and meat products. The exploitation group estimates that the potential market for the developed technology is considerable and have made modest estimations of 0.5% market penetration for meat processors SMEs and 2% for abattoirs within 5 years of completion of this project. A conservative projection of the market potential and market share for a 5-year period is summarised in the Deliverable 7.4 Final PUDF.

Impact on standards and policies
QMEAT will contribute to the development of new European quality standards, provided that it will allow establishing new objective and automatic classification methodologies based on the technological quality of the meat. This new technology will contribute to the general objectives of the European Union policy that promotes high level of food quality, animal health as well as animal welfare in order to improve the competitiveness of the European meat sector, and to promote the consumption of healthy meat products among European citizens.
Contribution to Community Society objectives
QMEAT will contribute to the continued growth of the European pigmeat sector as well as to increase the trust of European consumers by providing consortium SMEs, and eventual end-users, with a system that accurately determine the technological quality of the meat before processing, resulting in a better final product with regard to its sensory features. European consumer associates quality with taste and the perception of pork being healthy and safe. Meat quality is in turn associated to animal welfare. A study carried out by the EC reveals that 43% of the EU citizens consider animal welfare most or some of the time when purchasing meat . It has been demonstrated that the most important factor affecting meat quality is animal welfare rather than genetics .
The work programme, as well as the potential benefits of the project, will provide collaborative opportunities for end users SMEs of the consortium and standardize an important aspect of production. Consequently, QMEAT will be more effective and increase the knowledge base of consortium SME producers, provide them tools to expand their businesses, and support employment growth in the industry. Furthermore, QMEAT will assist in competing against tougher competition from non-EU countries and “New Wold” countries like USA, Brazil and Canada.
Main dissemination activities and exploitation of the results
During the project lifetime, the members have been actively involved in the dissemination activities with the aim of communicating and promoting the technical and commercial interest of the QMEAT technology to a wider audience. Each partner had a clear responsibility: the SMEs were primarily responsible for disseminating in trade fairs and to business contacts, whereas the RTDs performers were engaged in introducing the project to the scientific community during several conferences and workshops. During these events, technological and business non-confidential information have been spread between partners attending them.
In addition, to provide an attractive marketing image and help the dissemination of the project results outside the consortium, the coordinator ATEKNEA created dissemination materials such as a logo, a brochure and a poster that include non-confidential information used during the several events.
Furthermore, a project website has been created and has a restricted area for access and storage of technical information by consortium partners. In the public section, the members of the public can consult non-confidential information about the project, its benefits, its cost-effectiveness, but also relevant information of results. All the contents are supervised by the Exploitation Manager (JCB) to assess that only non-confidential information is published.

5.World Meat: FAPRI 2010 Agricultural Outlook / 333
7.S. Huete, B. Flach, EU-27, livestock and products annual report 2010. USDA Foreign Agricultural Service
8.Agricultural statistics: Livestock and meat, Eurostat 2008.
11.Confederation of the food and drink industries of the EU
12.CIAA Competitiveness Report 2010: Supporting the competitiveness of the European food and drink industry
13.D. J. O'Neill, P. B. Lynch, D. J. Troy, D. J. Buckley, J. P. Kerry, Influence of the time of year on the incidence of PSE and DFD in Irish pigmeat, Meat Science, Volume 64, Issue 2, June 2003, Pages 105-111.
14.CLITRAVI: Liaison Centre for the Meat Processing Industry in the European Union. FENAVIAN: Fédération nationale des fabricants de produits et conserves de viandes. AICE: Asociación de Industrias de la Carne de España.
15.EUROSTAT: Production and external trade of foodstuffs: Fresh meat and meat products. 2006 Sausage production
16.EUROSTAT: Number of enterprises, persons employed and turnover in the manufacturing of foodstuffs. 2007 Production, processing, preserving of meat, meat products

List of Websites: